Middle Class Without a Net: Savings,Financial Fragility,and Preferences Over Social Insurance

Abstract

In this article, we show that it is crucial to distinguish between liquid and illiquid wealth to understand how voters form preferences toward social insurance. Many households are financially fragile despite having high incomes and wealth, because they hold little liquid savings. We hypothesize, and show empirically, that this implies that a substantial group of voters show strong support for social insurance policies despite being wealthy and having high incomes, because of their limited ability to self-insure through own savings in case of an income shock. Our empirical analysis is based on a novel dataset from Denmark, which combines administrative data with high-quality measures of individual financial assets and survey measures of political preferences. Using data for other countries from the European Social Survey, we find evidence that our results hold more generally and are not specific to the Danish context.

Keywords

political economy social welfare programs public opinion

Introduction

Social insurance and redistribution are defining features of the welfare state, and a vast literature in political economy investigates the individual-level determinants of social policy preferences. Recent work has emphasized the importance of economic self-interest, including portability of skills (Iversen & Soskice, 2001), labor market risk (Rehm, 2009), wealth holdings (Ansell, 2014), and economic worries more generally (Hacker, Rehm, & Schlesinger, 2013), for understanding political preference formation. However, the focus on formal social insurance arrangements is incomplete—and recognized as such: In an influential analysis of social insurance and redistribution, Moene and Wallerstein (2001, p. 871) note that “[t]heoretically, the largest gap in our approach is the absence of a private alternative to publicly provided insurance. We have concentrated on the loss of income, a risk that cannot be insured privately.” But private supplements and alternatives to social insurance of income risks do exist, partly because formal social insurance programs often offer only partial income replacement and, in the case of unemployment insurance (UI), only for a limited time. Therefore, people’s reliance on formal social insurance is in practice widely supplemented by, and sometimes secondary to, informal self-insurance, of which the most important component is drawing on private savings (Gruber, 2001).

Yet, the availability of private savings and credit market opportunities differ considerably across individuals. The middle class is now becoming more financially fragile (Lusardi, Schneider, & Tufano, 2011) and heterogeneous (Atkinson & Brandolini, 2011). Many households with medium or high incomes are increasingly characterized by limited savings and accumulated debts, which leaves them “no margin for error in their financial lives” (Sullivan, Warren, & Westbrook, 2000, p. 250). Across industrialized countries, wealthy “hand-to-mouth” households—households that have positive illiquid wealth, typically from housing equity, but little or no liquid savings—constitute between 30% and 40% of the population (Kaplan, Violante, & Weidner, 2014). One consequence of this is that economic vulnerability, measured by a lack of access to economic buffers, is also common among middle-class and rich households and, thus, largely unrelated to current income. Following the literature on the economics of consumption and savings, we refer to people who are unable to access economic buffers to maintain current consumption as liquidity constrained.

In this article, we examine whether the presence of such liquidity constraints increases individual-level demand for social insurance and redistribution. Liquidity constraints is a core concept in economic theory, necessary to understand the link from current income to consumption. By introducing it into political science, the article adds considerable nuance to the concepts of income, wealth, and what it means to be “rich” in political science research. While existing research convincingly argues that perceived economic vulnerability matters for policy preferences (Hacker et al., 2013), we show that our conceptual unpacking of objective income and wealth measures are important for explaining people’s social policy preferences: Our key finding is that being liquidity constrained predicts support for increasing social insurance and redistribution, even among wealthy individuals. Liquidity constraints constitute a novel and undertheorized determinant of individual social policy preferences that cuts across the traditional fault line of class politics. Also, it is substantively relevant: In our data from Denmark, 40% of the survey population is liquidity constrained, a number comparable to that found in the U.S. and other industrialized countries (Kaplan et al., 2014).

The welfare state has two distinct functions: redistribution, which serves to equalize the income distribution, and social insurance, which provides a financial cushion against misfortune. The classical accounts of the welfare state (e.g., Esping-Andersen, 1990; Esping-Andersen & Korpi, 1984; Korpi, 1983) focused on redistribution and how the distribution of power resources—including, but not limited to, relative income and wealth—shaped welfare state trajectories. Focusing on individual-level preferences, Meltzer and Richard (1981) showed that citizens with a relatively low income are more likely to favor redistribution and support the welfare state than are the more well-to-do, and, recently, Ansell (2014) showed that rising housing wealth is associated with lower support for redistribution. A different strand of the welfare state literature has, instead, focused more on social insurance and how attitudes are shaped by risk coalitions (Baldwin, 1990) based on relative vulnerability in the labor market (e.g., Cusack, Iversen, & Rehm, 2006; Estevez-Abe, Iversen, & Soskice, 2001; Iversen & Soskice, 2001; Moene & Wallerstein, 2001, 2003; Rehm, 2009, 2011) and economic worries (Hacker et al., 2013).

Our political economy approach combines the two dimensions of wealth and vulnerability and focuses on the ability to weather economic shocks. The argument has two parts: First, the middle class and the rich, despite being by definition wealthy and/or having high incomes, can be economically vulnerable. In part, this is because they have pre-committed expenditures that are costly to adjust, such as school payments, mortgages, or general debt servicing (Chetty & Szeidl, 2007; Warren & Tyagi, 2003). Second, the impact of wealth on vulnerability critically depends on how liquid the wealth is: Illiquid wealth, including housing wealth, is at best an imperfect instrument for smoothing consumption when faced with adverse economic shocks. This means that to understand the relationship between economic resources and social policy preferences, we need to carefully distinguish not only income and wealth but also the composition of wealth by its liquidity or, in other words, by its usefulness as short-term income insurance. Our core hypothesis is that liquidity constraints explain attitudes to social insurance for all levels of illiquid wealth: Despite being middle-class or rich by standard metrics, liquidity constrained households should share the attitudes of the poor on issues of social insurance policy.

Our primary analyses combine two types of data rarely available together: a large-scale survey of political attitudes and preferences combined with unusually rich and detailed administrative data containing information on liquid and illiquid savings, labor market history, and demographics. The level of detail enables us to distinguish individuals along a number of important dimensions and to assess competing and complementary hypotheses in the political economy of social policy preference formation in a unified empirical framework. Our key findings obtain for different measures of liquidity constraints, including preexisting bank debt and the price of borrowing. We show that our findings are robust to accounting for the main alternative explanations suggested by the recent political economy literature (all of which also find empirical support).

We extend the main analysis in two directions. First, we account for the fact that savings and liquidity constraints are not exogenously given at the individual level and may both reflect future adverse circumstances beyond those captured by risks of unemployment, including the existence of social insurance itself (Engen & Gruber, 2001) as well as more stable traits, including, for example, risk aversion, impatience, or a preference for thrift. We utilize individual-level data on savings behavior more than a decade before the survey to assess whether current liquidity constraints reflect contemporaneous income shocks or more stable characteristics. Our results suggest that both sources of liquidity constraints are important, even if the latter seems to weigh more. As a second extension, we provide a first pass at generalizing our findings to a cross-country setting. This is important to provide external validity to the one-country study and, also, because all advanced economies have significant shares of liquidity constrained individuals. Despite different credit market structures and welfare state arrangements, we find—using data from the European Social Survey—that liquidity constraints predict social policy preferences across 22 countries.

To the best of our knowledge, this is the first article to consider the political consequences of liquidity constraints, even if they have long been a key concept in economics. It is well documented, for example, that liquidity constraints exacerbate households’ vulnerability to income and employment shocks: When households hold little liquid wealth or are unable to borrow, a negative income shock leads to falling consumption and living standards (e.g., Zeldes, 1989). In the context of labor markets, recent evidence suggests that liquidity constraints tend to reduce the duration of job search for unemployed households (Basten, Fagereng, & Telle, 2014; Card, Chetty, & Weber, 2007). Such results are consistent with the argument presented in this article: Liquidity constrained households have preferences for (increased) unemployment benefits, because such benefits allow them to sustain living standards in the event of job and income loss and, also, because they allow them more search time to find a job that matches their skills.

The remainder of the article proceeds as follows. In the next section, we present the theory and the predictions that guide the empirical analyses. The “Context” section presents the institutional context, and the “Data and Estimation Approach” section describes the data and our empirical approach. The “Results” section presents the main results, with additional robustness analyses in the Online Appendix (OA), and the “Further Perspectives” section briefly introduces cross-country findings as well as our analysis of the relative importance of stable traits and idiosyncratic shocks for savings behavior and preference formation. The final section concludes and proposes avenues for future research.

Theoretical Framework

The political economy literature on the formation of social policy preferences argues that individuals’ preferred policies vary with their economic situation. This literature has focused almost exclusively on the level of economic determinants such as income and, to a lesser extent, wealth including housing and human capital. The liquidity of these economic determinants is an almost entirely neglected, but equally important, dimension of the economics of individual social policy preferences.¹ In this section, we combine economic theories of consumption, savings, and insurance with the political economy of redistribution and social insurance to arrive at testable hypotheses linking different types of wealth and savings to individual social policy preferences.

The basic assumption of the canonical economic model of consumption and savings—the permanent income hypothesis (Friedman, 1957)—is that people prefer a stable consumption flow to one that fluctuates. To maintain a constant standard of living, individuals therefore divide their lifetime resources evenly over time. This implies that a transitory drop in income should not translate into an equal drop in living standards. Instead, people cushion the shock by borrowing or by running down liquid assets such that living standards fall marginally for many periods rather than precipitously for a few periods. Yet, individuals who have limited liquid savings relative to income, or who face a high interest rate in the borrowing market, are liquidity constrained (Zeldes, 1989) and unable to cushion shocks and sustain consumption during periods of lower income. Because social insurance allows people who are liquidity constrained to realize a consumption path closer to their pre-shock level, we hypothesize that liquidity constrained households or individuals will favor unemployment protection and social insurance in general. Individuals thus compensate for a limited ability to self-insure by demanding increased social insurance. In that sense, social insurance works as a substitute for private insurance.²

Even if it is an axiomatic assumption in intertemporal models of consumption that people dislike variation and prefer constant living standards, there is evidence that people’s savings do not adequately prepare them for income shocks. Focusing on job risk, Gruber (2001) and Basten, Fagereng, and Telle (2016) find that while some households are adequately prepared, most are not. Consistent with this, Hacker et al. (2013) report considerable worries over the adequacy of savings among U.S. citizens. The inability to smooth consumption in the event of an income shock and financial worries is not (only) a result of low income and is not confined to the poor: liquidity constraints extend to the upper part of the income distribution in the form of limited savings and high credit costs.

To understand how one can be wealthy and financially vulnerable at the same time, it is important to distinguish between liquid and illiquid wealth. An individual’s financial wealth consists of some assets that are liquid (e.g., cash and bank deposits) and some that are illiquid (e.g., real estate). Liquid assets are better suited than illiquid assets as a means of self-insurance against short-term shocks to income or employment, as illiquid assets can be traded only at a (sometimes substantial) transaction cost or at a considerable delay. Illiquid assets are therefore more appropriate for transferring wealth across the life cycle (from working age to retirement, for example, the case considered by Ansell, 2014) than to mitigate the effects of temporary shocks.³ Indeed, the main illiquid asset of most households, real estate, is rarely liquidated during unemployment (Chetty & Szeidl, 2007). This suggests the existence of a group of “vulnerable rich”—characterized by high incomes and substantial illiquid wealth, but little liquid wealth with which to weather shocks and smooth consumption. Because illiquid wealth is an imperfect substitute for liquid wealth as a means for short-term self-insurance, we hypothesize that liquidity constraints predict social policy preferences for all levels of illiquid wealth.

This hypothesis implies that self-interested support for the welfare state can extend into and beyond the middle class. It suggests a new fault line in welfare state politics between those who are able to self-insure and those who have to rely more on social insurance. Importantly, these groups are different from the familiar rich-poor dichotomy, which brings us to another point about the importance of distinguishing between liquid and illiquid wealth: Whereas liquidity constraints are prevalent across the income distribution, illiquid wealth is essentially a proxy for income: High-income individuals tend also to be wealthy in terms of illiquid wealth. Thus, if one does not distinguish between liquid and illiquid assets, one will simply overlook that financial vulnerability can also characterize middle-class and high-income individuals. In summary, the distinction between liquid and illiquid wealth is important in analyses of the political economy of social policy preferences both because liquid wealth serves an insurance purpose against short-term risk for which illiquid wealth compensates poorly, and because the distribution of liquidity constraints cuts across the traditional fault line of relative income, whereas the distribution of illiquid wealth follows more closely the traditional rich-poor divide.

While our main theoretical focus is on preferences over social insurance, these are in practice difficult to separate from preferences over redistribution. The reason is that the two functions of the welfare state often overlap in practice. The social insurance dimension of the welfare state serves as a safeguard against misfortune and includes unemployment protection, health care, and disability benefits, while the redistribution dimension serves to equalize the income distribution through (progressive) taxation of earnings and assets. However, there is an element of insurance in redistribution (by, e.g., providing a lower bound beneath which living standards cannot fall, despite misfortune), and an element of redistribution in UI (since it is often partly financed by progressive taxation and payments often are linked to pre-unemployment earnings). As well, the two concepts are often employed interchangeably in analyses of social policy preferences.⁴ In our empirical analysis, we therefore show that liquidity constraints predict attitudes to redistribution as well as to social insurance. We focus on results on social insurance in the main article and present most results on redistribution in the OA.

Context

We study the importance of liquidity constraints for social policy preferences in the context of Denmark. The welfare system in Denmark is considered to be comparatively generous (e.g., Kautto, Fritzell, Hvinden, Kvist, & Uusitalo, 2001). The UI system consists of a baseline means-tested social transfer that everyone is entitled to and a higher transfer available to members of a UI fund. Membership of the UI fund is voluntary. Two thirds of the costs are covered by the members who pay a flat fee. The rest is financed by general taxes, which means that the UI system includes an element of redistribution. Eligibility for unemployment benefits from the UI fund requires at least 1 year’s membership and no less than 6 months of full-time employment over a period of 3 years. Benefits amount to 90% of pre-unemployment wages up to a threshold of about US$35,000 per year, roughly equivalent to the median income, and is capped beyond that. The maximum duration of UI is 2 years, after which the worker can shift to the lower means-tested social transfer with an unlimited duration.⁵

The share of liquidity constrained individuals in our data, defined by liquid assets relative to income, is around 40%. This is similar to the numbers estimated for a number of advanced industrialized countries, including Australia, Canada, France, Germany, Italy, and Spain, the United Kingdom, and the United States (Kaplan et al., 2014). For the average Danish household, most of its wealth is held in housing, as Denmark has a rate of homeownership of around 56% (Statistics Denmark). This is similar to many advanced countries. This composition of wealth and the fact that homeownership is often financed through high monthly mortgages that deplete the stock of liquid wealth and are costly to adjust create a vulnerable middle class: They may be wealthy in terms of total assets and income but are poor in terms of liquid savings.

Data and Estimation Approach

Data Sources

For the empirical analysis, we combine two unique data sources. To obtain information about policy preferences, we use data from a large survey of around 6,000 respondents each year over the years 2010 to 2013. The respondents are randomly drawn from the Danish working-age population with some labor market participation in the period 1998 to 2004. We use the observation of the first interview and exclude people who are permanently out of the labor force, yielding around 8,000 unique respondents.⁶ To obtain information about personal finances, labor market conditions, and demographics, we combine the survey data with uniquely detailed data from Statistics Denmark covering the entire Danish population back to 2000. This is linked to the survey data via a personal identification number assigned to all Danish residents.⁷ The administrative data from Statistics Denmark is reported from third parties; for example, information on earnings is reported by employers; information on transfer income is reported by government agencies; information about the value of assets and liabilities is reported directly by banks and other financial institutions.⁸ This means both that our variables are generally very close to the “true” values and that there are essentially no missing items. Summary statistics are available in Table A.24 in the OA.

Outcome Variables

We focus on two outcomes: preferences for UI and for redistribution. Regarding UI, respondents are asked whether they think that the state should do more to support those who become unemployed, and they can answer on a three-point scale: “less than today,” “unchanged,” or “more than today.” For redistribution preferences, the respondents are asked to place themselves on a scale from one to five, where one is “the government should do everything it can to raise living standards of the poor” and five is “it is not the government’s responsibility, people should take care of themselves.” We normalize both outcomes to run on a scale from zero to one such that coefficients are comparable across outcomes.

Liquidity Constraints and Other Explanatory Variables

People are liquidity constrained if they are unable to access liquid funds on short notice. Liquidity constraints result in an inability to smooth consumption and sustain living standards during a temporary income shock. Conceptually, we are interested in the extent to which people can sustain their current living standard if, for whatever reason, they did not receive their paycheck. We therefore consider liquid assets relative to income and use as our main independent variable a measure of liquidity constraints that is standard in the economics literature: An individual is considered liquidity constrained if the sum of her liquid assets is less than one and a half month of disposable income (e.g., Johnson, Parker, & Souleles, 2006; Leth-Petersen, 2010; Zeldes, 1989). The value of liquid assets is equal to the sum of bank deposits, the value of stocks, and the value of bonds. It is measured on December 31 and reported by banks to tax authorities for tax purposes. None of our findings hinge on the specific operationalization of liquidity constraints, and we present results below and in the OA for a range of alternative measures for liquidity constraints, including different cutoffs (at the sum of liquid assets less than 1 month and 2 months of disposable income, respectively), the fraction of years liquidity constrained since year 2000 (which captures a more permanent status), and a measure of the individual marginal cost of liquidity raised in the credit market (Kreiner, Lassen, & Leth-Petersen, 2019).⁹

As controls, we include determinants of social policy preferences known from the literature. We show robustness to inclusion of different detailed measures of income, and as our main income control variable, we use a measure of “permanent income” equal to the individual’s average annual gross income since 2000.¹⁰ We furthermore include the individual’s perceived unemployment risk, a measure of skill specificity that captures transferability of skills across jobs (the measure SI from Iversen and Soskice (2001)), and a measure of illiquid wealth. Our measure of illiquid wealth focuses on home equity, which is defined by the public assessment of the value of all owned homes minus the total value of mortgage debt.¹¹ All these variables are based on information from the administrative registers, except for the perceived unemployment risk, which is based on a survey question where the respondent is asked how likely it is that she will experience a period of unemployment within the next year. This measure is a strong predictor of the realized unemployment next year (which we observe in the administrative data) and is highly correlated with the history of unemployment and unemployment rates within the respondent’s industry and occupation (Alt, Barfort, & Lassen, 2017; Hendren, 2017).

As additional controls we include information from the administrative registers about demographics (age, marital status, children, immigrant status, gender), current labor market status, membership of a UI fund, and years of education. We also include a measure of whether respondents consider success to be mostly due to hard work or luck and connections. This variable captures respondents’ beliefs about fairness and deservingness, which has been shown to be a strong determinant of social policy preferences (e.g., Alesina & Angeletos, 2005).

Estimation

To examine the association between liquidity constraints and the outcomes of interest, we estimate the following equation using ordinary least squares (OLS):

Y_{i} = α + δ C_{i} + X_{i} β + ε_{i} .

(1)

Here, Y_i represents the social policy preferences of individual i, C_i is the dummy for being liquidity constrained for i, and X is a vector of controls. In some of the specifications, we extend the model with an interaction term to examine the interplay between liquid and illiquid wealth. In the penultimate section, we implement a two-step procedure to split up the more persistent and the more transitory part, respectively, of liquidity constraints. Both alternative models will be explained in greater detail. Robust standard errors are reported throughout.¹²

Results

Income, Assets, and Liquidity

A central reason for introducing the concept of liquidity constraints to political economy is that it is distinct from measures of income. Liquidity constraints are prevalent across the income distribution. This is documented in Figure 1, which shows a binned scatter plot of the share of liquidity-constrained individuals in bins of gross income. While there is a weak downward trend, the share of liquidity constrained individuals stays high at around 40% to 50% for most of the income distribution.

Figure 1.

Liquidity and income.

This pattern is partly due to the fact that much consumption is “pre-committed” and difficult to adjust from period to period (Chetty & Szeidl, 2007; Warren & Tyagi, 2003). This type of consumption includes rent, mortgages, other servicing of loans, utility bills, membership fees, tuition, and alimony expenditures. These expenditures must be incurred regularly unless one pays a, sometimes substantial, transaction cost to modify the existing commitments, and they are usually increasing in income: For instance, people with higher incomes buy more expensive houses and face higher mortgages. This means that also in the upper part of the income distribution, many households deplete their holdings of liquid assets each month and would not be able to maintain their living standard for even short periods of time if they temporarily stopped getting their paycheck. Figure 1 is consistent with the cross-country findings of Kaplan et al. (2014) and with Lusardi et al.’s (2011) finding that a sizable proportion of the American middle class considers itself to be financially fragile.

In contrast, illiquid wealth, which clearly constitutes the bulk of overall wealth, is highly correlated with income. This is documented in Figure 2. For larger income shocks and for consumption smoothing over a longer time horizon, from working age to retirement for instance, high-income individuals therefore have more options: if an income shock becomes sufficiently large, they can adjust a high level of committed expenditures by downsizing or taking out a home equity loan. In the regression analysis below, we consider separately the distinct effects of liquid and illiquid wealth on social policy preferences; for now, note from Figures 1 and 2 that the distinction between liquid and illiquid wealth matters for how people are divided across the income distribution. Liquidity constraints cut across income groups, while illiquid wealth follows a traditional rich–poor categorization.

Figure 2.

Home equity and income.

Liquidity Constraints and Social Policy Preferences

Table 1 reports coefficients from OLS regressions of preferences for UI on liquidity constraints, corresponding to Equation 1. The first column shows the unconditional association between liquidity constraints and social policy preferences, while control variables and various measures of income are included in the columns that follow.

Table 1.

Liquidity Constraints and Social Insurance Preferences.

Variables	(1)	(2)	(3)	(4)	(5)
Variables	More UI	More UI	More UI	More UI	More UI
Liquidity constrained	0.027***	0.032***	0.031***	0.030***	0.048***
	(0.006)	(0.006)	(0.006)	(0.006)	(0.015)
Gross income (DKK 100,000)			−0.012***
Gross income (DKK 100,000)			(0.002)
Permanent income (DKK 100,000)				−0.017***	−0.016***
Permanent income (DKK 100,000)				(0.002)	(0.002)
Permanent Income × Liquidity Constrained					−0.005
					(0.004)
Controls
Labor market
Unemployed		0.059***	0.044**	0.047***	0.046***
		(0.017)	(0.017)	(0.017)	(0.017)
Self-employed		−0.071***	−0.077***	−0.070***	−0.070***
		(0.012)	(0.012)	(0.012)	(0.012)
Outside the labor force		0.045***	0.027**	0.030**	0.030**
Outside the labor force		(0.013)	(0.014)	(0.014)	(0.014)
Unemp insurance		0.036***	0.036***	0.035***	0.035***
		(0.008)	(0.008)	(0.008)	(0.008)
Demographics
Age		0.002***	0.002***	0.003***	0.003***
		(0.000)	(0.000)	(0.000)	(0.000)
Woman		0.044***	0.033***	0.027***	0.027***
		(0.006)	(0.006)	(0.006)	(0.006)
Single		0.036***	0.035***	0.035***	0.035***
		(0.008)	(0.008)	(0.008)	(0.008)
Immigrant		0.020	0.014	0.012	0.011
		(0.017)	(0.017)	(0.017)	(0.017)
Children		−0.012*	−0.007	−0.003	−0.003
		(0.007)	(0.007)	(0.007)	(0.007)
Short educ		−0.013	−0.010	−0.009	−0.009
		(0.008)	(0.008)	(0.008)	(0.008)
Medium educ		0.004	0.013	0.015	0.015
		(0.009)	(0.009)	(0.009)	(0.009)
Long educ		−0.012	0.011	0.014	0.015
		(0.011)	(0.011)	(0.011)	(0.011)
Attitudinal
Success due to effort		−0.052***	−0.050***	−0.049***	−0.049***
Success due to effort		(0.006)	(0.006)	(0.006)	(0.006)
Observations	8,035	8,035	8,035	8,035	8,035
R ²	.003	.047	.052	.054	.054

The table reports coefficients from OLS regressions of respondents’ attitudes to unemployment insurance on a dummy for being liquidity constrained and the controls shown. All models include a constant term and dummies for the survey years (not reported to save space). Robust standard errors in parentheses. UI = unemployment insurance; OLS = ordinary least squares.

***

,**, and * indicate significance at the 1%, 5%, and 10% levels.

Throughout, the coefficient on the indicator for being liquidity constrained is significant and substantial. This holds unconditionally and after controlling for various measures of income, detailed data on the individuals’ labor market situation, and demographics: An individual who is liquidity constrained is three percentage points more likely to support a more generous UI system than is a nonconstrained individual who is similar in terms of covariates. To illustrate the substantive relevance of liquidity constraints for understanding social policy preferences, the coefficient is similar in size to those estimated for women and respondents outside the labor force. Table 1 thus confirms that private savings and public social insurance are substitutes. As expected, permanent income is significantly negative throughout the analysis: Individuals with higher incomes are less welcoming of governments’ programs and interventions. This shows that the coefficient on liquidity constraints does not capture (nor is it captured by) a more traditional rich-versus-poor explanation of welfare state preferences.

Similar conclusions obtain for redistributive preferences, presented in Table A.1. The point estimate for liquidity constraints is consistently significant, but smaller in size than for social insurance. Even if the difference in coefficient magnitude is not significant, it is worth noting. Theoretically, we expect liquidity constraints to shape demand for social insurance. We expect liquidity constraints to predict redistributive preferences only because redistribution also contains an element of insurance. Tables 1 and A.1 show that liquidity constraints are more strongly correlated with preferences for social insurance than for redistribution, and we show below that alternative measures of liquidity constraints (the marginal cost of borrowing and the existence of bank debt) are strongly associated with demand for social insurance but uncorrelated with redistributive preferences. This is consistent with the asset model from Iversen and Soskice (2001) wherein the level of an economic asset determines preferences for redistribution, while its liquidity determines preferences for social insurance. Iversen and Soskice focus on human capital, but the results in Tables 1 and A.1 (and most clearly in Table 3 where we use alternative measures of liquidity constraints) suggest that the theory may also apply for financial wealth in that liquidity constraints are more strongly associated with attitudes to social insurance than to redistribution.

Comparing the coefficients on the control variables in Tables 1 and A.1 also suggest a meaningful distinction between the redistributive and the insurance functions of the welfare state. Respondents who are at risk economically are generally in favor of social insurance: Respondents who are unemployed, outside the labor force, members of a UI fund, or single have strong preferences over the generosity of the UI system but do not have strong views on redistributive policies. Conversely, the level of education is correlated with redistributive preferences but not with preferences for social insurance. This is also consistent with Iversen and Soskice (2001).

Finally, the rightmost columns show results for the interaction of income and being liquidity constrained. Since the replacement rate is lower for high-income individuals, due to the cap on unemployment benefits, they would have less reason to support the existing UI system. However, the survey question asked individuals if they think more should be done to help those becoming unemployed, and this could conceivably include extending coverage and replacement rates. The lack of significance of the interaction term supports this latter interpretation and is consistent with our main argument that social insurance is important also for high-income earners.

Competing Explanations for Social Policy Preferences: Liquid Wealth, Illiquid Wealth, and Labor Market Risk

We now examine whether the association between liquidity constraints and social policy preferences changes when the model is extended to include home equity (Ansell, 2014), skill specificity (Iversen & Soskice, 2001), and perceived unemployment risk (Rehm, 2016). Each additional variable is included one at a time: first in a simple model that only includes the liquidity constraint-dummy and the new variable, then in a more complete model that includes the full set of controls from Table 1 and, finally, in a model that includes all controls and the three new determinants. Table 2 reports the results for social insurance preferences, and Table A.2 shows results for preferences over redistribution.

Table 2.

The Effects of Liquidity Constraints, Wealth, and Labor Market Risk on Social Insurance Preferences.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Variables	More UI	More UI	More UI	More UI	More UI	More UI	More UI
Liquidity constrained	0.020***	0.028***	0.027***	0.030***	0.026***	0.030***	0.028***
	(0.006)	(0.006)	(0.006)	(0.006)	(0.006)	(0.006)	(0.006)
Permanent income (DKK 100,000)		−0.016***		−0.017***		−0.016***	−0.015***
		(0.002)		(0.002)		(0.002)	(0.002)
Home equity (DKK 100,000)	−0.0018***	−0.001***					−0.001***
	(0.000)	(0.000)					(0.000)
Skill specificity			0.000	0.001			0.001**
			(0.000)	(0.000)			(0.000)
Unemployment risk (%)					0.001***	0.001***	0.001***
					(0.000)	(0.000)	(0.000)
Controls
Labor market		✓		✓		✓	✓
Demographics		✓		✓		✓	✓
Attitudinal		✓		✓		✓	✓
Observations	8,035	8,035	8,035	8,035	8,035	8,035	8,035
R ²	.007	.055	.003	.054	.010	.057	.058

The table reports coefficients from OLS regressions of respondents’ attitudes to unemployment insurance on a dummy for being liquidity constrained and the controls shown. The set of labor market controls, demographic controls, and attitudinal controls not reported are the same as in Table 1. All models include a constant term and dummies for the survey years. Robust standard errors in parentheses. UI = unemployment insurance; OLS = ordinary least squares.

***

,**, and * indicate significance at the 1%, 5%, and 10% levels.

The coefficient on the dummy for liquidity constraints remains significant and substantial across specifications. Compared with the results in Table 1, the estimated association remains unchanged when the other important determinants of social policy preferences are controlled for. This confirms that liquidity status is an important input into the process of social policy preference formation, and that this is not captured by income, illiquid wealth, labor market risk, or skill composition. Note that the three determinants that were added to the model all have the expected sign and are significantly related to preferences for UI. Comparing the coefficients in Table 2 and A.2 shows that skill specificity and perceived unemployment risk are more strongly associated with attitudes to social insurance than redistribution, whereas the opposite holds for home equity. We also interacted liquidity constraints with the risk of unemployment to see whether people who are more at risk in the labor market are more supportive of social insurance when they are less able to self-insure, but found no significant effects (see Tables A.13-A.15 in the OA for different measures of unemployment risk and different measures of liquidity constraints, which all yield the same conclusion).¹³

Thus far, we have identified the distinct effects of having limited access to liquid assets and illiquid wealth in the form of home equity. To understand in more detail the relationship between liquid and illiquid wealth in shaping individual social policy preferences, we examine the interaction between the two in a model that is otherwise identical to column (7), Table 2. Figure 3 shows the result graphically, and Tables A.18 and A.19 in the OA show the corresponding regression tables.

Figure 3.

Effect of liquidity constraints on support for UI over values of home equity.

Consistent with expectations, Figure 3 shows that liquidity constraints are significantly associated with demand for social insurance for almost all levels of illiquid assets. This confirms that, as a means of self-insurance against short-term risks, illiquid wealth is an imperfect substitute for liquid wealth. The coefficient on liquidity constraints remains sizable and significant at the 95% level until the rightmost part of the illiquid wealth distribution (around the 90th percentile). While home equity may serve as a buffer against large and persistent shocks, it does not fulfill the need for insurance that comes from a lack of liquid wealth holdings. As noted above, the reason for this is that the cost of accessing illiquid wealth is often substantial relative to the income loss caused by temporary economic hardship. However, if such costs are constant regardless of the size of, say, a secondary mortgage, or the gains from liquidity are higher relative to the costs for high-wealth households, the costs of transforming illiquid wealth into liquid wealth will be relatively smaller for more wealthy individuals. This would produce a relationship between liquidity constraints, illiquid wealth, and support for UI as the one shown in Figure 3.¹⁴

Finally, Table 3 presents the full specifications of Table 2 for three alternative measures of liquidity constraints. The first two columns use the marginal interest rate, that is, the interest rate on the borrowing market currently faced by survey respondents (see OA II.I for details on how this has been measured). This constitutes a proxy, and presumably lower bound, for the cost of additional credit. Columns (3) and (4) use an indicator for having bank debt, and columns (5) and (6) show results for the fractions of years over a decade preceding the survey that respondents were liquidity constrained. Across measures, we find strong effects of the lack of economic buffers on preferences for social insurance, but less clear results for attitudes toward redistribution. One possible reason is that exposure on the borrowing market is a strong signal of lack of insurance possibilities for both the rich and the poor, two groups who have different preferences over redistribution. The findings underscore, again, that social insurance and redistribution are distinct functions of the welfare state and that survey respondents are able to distinguish meaningfully between them.¹⁵

Table 3.

Alternative Measures of Liquidity Constraints.

Variables	(1)	(2)	(3)	(4)	(5)	(6)
Variables	More UI	More redistribution	More UI	More redistribution	More UI	More redistribution
Borrowing market interest rate	0.002***	0.001
	(0.0004)	(0.0004)
Bank debt			0.020***	0.004
			(0.007)	(0.007)
Share years liquidity constrained					0.066***	0.049***
					(0.009)	(0.009)
Permanent income (DKK 100,000)	−0.015***	−0.013***	−0.016***	−0.013***	−0.015***	−0.012***
	(0.002)	(0.002)	(0.002)	(0.002)	(0.002)	(0.002)
Home equity (DKK 100,000)	−0.001***	−0.001***	−0.001***	−0.001***	−0.001**	−0.001***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Skill Specificity	0.001**	−0.001	0.001**	−0.001	0.001**	−0.001
	(0.000)	(0.001)	(0.000)	(0.001)	(0.000)	(0.001)
Unemployment risk (%)	0.001***	0.0002**	0.001***	0.0002*	0.001***	0.0002*
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Controls
Labor market	✓	✓	✓	✓	✓	✓
Demographics	✓	✓	✓	✓	✓	✓
Attitudinal	✓	✓	✓	✓	✓	✓
Observations	7,005	7,005	8,035	8,035	8,035	8,035
R ²	.0551	.0708	.0583	.0732	.0031	.0732

The table reports coefficients from OLS regressions of respondents’ attitudes to unemployment insurance and redistribution on three different measures of liquidity constraints and the controls shown. The set of labor market controls, demographic controls, and attitudinal controls not reported are the same as in Tables 1 and 2. All models include a constant term and dummies for the survey years. Robust standard errors in parentheses. UI = unemployment insurance; OLS = ordinary least squares.

***

,**, and * indicate significance at the 1%, 5%, and 10% levels.

Further Perspectives

In this section, we explore the relative importance of temporary shocks and more permanent characteristics in explaining when people are liquidity constrained and what that means for our results. We, also, briefly address the generalizability of our findings in a cross-section of countries.

The presence of liquidity constraints is not randomly distributed across individuals. Liquidity constraints may reflect unanticipated economic shocks or individual traits such as impatience, limited risk aversion, or personality. To see whether liquidity constraints and social policy preferences are jointly determined by individual traits, we control for measures of personality and spending temperament. Tables A.20 and A.21 include an indicator for respondents who report finding it difficult to control expenses; Tables A.22 and A.23 control for Big Five personality traits using a 10-item scale. Conditioning on personality does not, however, change the size or the significance of the coefficient on the dummy for being liquidity constrained.

As an alternative approach, we try to disentangle the relative importance of stable traits and idiosyncratic shocks in determining the existence of liquidity constraints and, through liquidity constraints, social policy preferences. We begin by looking at historical patterns of savings behavior at the individual level. Figure 4 shows the distribution of the number of years an individual in our sample was liquidity constrained over the period 2000 to 2012. The figure reveals large variation at the individual level—ranging from some that are never constrained to some that are always constrained. This is consistent with liquidity constraints sometimes being forced upon people by shocks such as unemployment and sometimes resulting from individual characteristics that are more stable. For many people liquidity constraints is a relatively persistent or recurrent condition.

Figure 4.

The number of years people have been liquidity constrained, 2000-2012.

In Table 4, we try to determine the relative importance of liquidity constraints arising from shocks and persistent characteristics, respectively, for political attitudes. The idea is that if liquidity constraints in the years of the surveys, 2010 to 2013, are correlated with financial characteristics going back more than a decade, this suggests that liquidity constraints are at least partly the result of factors that are too persistent to be explained by what is usually thought of as a transitory shock.¹⁶

Table 4.

Persistence and Shocks in the Effect of Liquidity Constraints.

Variables	(1)	(2)	(3)
Variables	Liquidity constrained	More UI	More redistribution
Liquidity constrained (predicted)		0.081***	0.050*
Liquidity constrained (predicted)		(0.028)	(0.027)
Liquidity constrained (residual)		0.021**	0.010
Liquidity constrained (residual)		(0.008)	(0.008)
Liquidity constrained in 2000	0.280***
	(0.014)
Permanent income (DKK 100,000)	−0.012**	−0.017***	−0.008***
	(0.005)	(0.003)	(0.003)
Home equity (DKK 100,000)	−0.004***	−0.001*	−0.001
	(0.001)	(0.001)	(0.001)
Skill specificity	−0.000	0.001	−0.001**
	(0.000)	(0.000)	(0.000)
Unemployment risk (%)	−0.000	0.0004***	0.0003**
	(0.000)	(0.0001)	(0.0001)
Controls
Labor market	✓	✓	✓
Demographics	✓	✓	✓
Attitudinal	✓	✓	✓
Observations	7,951	7,951	7,951
R ²	.159	.060	.074

The table reports coefficients from OLS regressions. The set of labor market controls, demographic controls, and attitudinal controls not reported are the same as in Tables 1 to 3. All models include a constant term and dummies for the survey years. Robust standard errors in parentheses. UI = unemployment insurance; OLS = ordinary least squares.

***

,**, and * indicate significance at the 1%, 5%, and 10% levels.

In column (1), we regress the presence of liquidity constraints in the years of the survey on an indicator for being liquidity constrained in 2000 and our usual battery of controls; this shows that being liquidity constrained a decade before the survey is a strong predictor of being liquidity constrained at the time of the survey. We use these estimates to predict current liquidity constraints. In the next step, we use predicted constraints as a regressor in analyses of social policy preferences alongside the residual variation in the condition of being liquidity constrained; this residual variation reflects the influence of individual-level economic shocks not captured by the more permanent traits identified in the analysis in column (1). This addresses the specific challenge to identification that idiosyncratic shocks might jointly determine the condition of being constrained as well as political attitudes. In columns (2) and (3) of Table 4, the condition of being liquidity constrained predicted from the presence of liquidity constraints in 2000 is significant and the estimate is larger in magnitude than in the basic estimation presented in the preceding tables. These results suggest that fixed or very persistent factors and traits that govern savings behavior play an important role in the condition of being liquidity constrained and, through liquidity constraints, also in shaping people’s political attitudes (see also Duch & Rueda, 2016). At the same time, the transient component in the condition of being liquidity constrained is also significant and confirms the importance of current economic circumstances in explaining liquidity constraints and, through these, individual social policy preferences.

We turn now to the generalizability of our findings. Across countries, substantial shares of the population are liquidity constrained, including groups with home equity (Kaplan et al., 2014). It is natural, therefore, to ask whether our detailed findings from Denmark generalize to a broader group of countries. As described in the “Context” section, the level of social insurance provided by the government in Denmark is comparatively high. A priori one would expect that the imperfections in the private market would be more important in settings with less generous welfare programs: The lower the level of public insurance, the more important is the ability to self-insure.

While we do not have cross-country data with the same level of detail that we have for Denmark, we use the most suitable proxy for liquidity constraints and a measure of social policy preferences from the European Social Survey. We use Round 5, from 2010, and define a person to be liquidity constrained if the respondent reports finding it difficult or quite difficult “to borrow money to make ends meet.”¹⁷ As the outcome variable, we use a question on whether “the government should take measures to reduce differences in income levels,” which runs on a 5-point scale from strongly agree to strongly disagree. For each country separately, we regress the outcome variable on our dummy for being liquidity constrained and a set of control variables (see notes to Figure 5). The country-specific coefficients on liquidity constraints are shown in Figure 5 together with confidence bounds (left axis). At the same time, we plot the country-specific compensation rate as a measure of the generosity of the national UI system (right axis).

Figure 5.

Liquidity constraints and social policy preferences across countries.

For 19 out of the 22 countries, the coefficient is positive, and for 14 countries it is significantly different from zero at the 95% level. As with the Danish data, the correlation is substantial: For most countries, the coefficient is larger than the coefficient on the gender dummy and, remarkably, larger than the coefficient on the dummy for having been unemployed within the last week before the interview. The coefficient estimated for Denmark is below the median, confirming our presumption that liquidity constraints would have a smaller effect here relative to many other developed countries. The countries are ranked according to the estimated coefficient, and the trend line for unemployment compensation rates indicates that it is in fact the case that liquidity constraints matter more in countries with a less generous welfare system: If misfortune hits and the level of social protection is limited, the ability to self-insure becomes all the more important.¹⁸ This suggests that the results from Denmark, a generous Nordic welfare state, may in fact represent something of a lower bound on the role of liquidity constraints, and that our results are likely to hold a fortiori in countries with less generous welfare systems (including the U.S. which is not a part of the ESS study). A plausible corollary of this finding is that the number of liquidity-constrained households should vary in tandem with the generosity of the welfare state: In less generous welfare states, to substitute for the limited availability of social insurance, there will be more informal self-insurance and, thus, fewer liquidity constrained individuals or households and vice versa in the more generous welfare states. This also suggests the existence of two different self-sustaining political equilibria in welfare politics: one with generous social insurance and many liquidity constrained, and one with less generous social insurance and few liquidity constrained. While we do not, at present, have the data to test this hypothesis, it provides an interesting avenue for future research.

Conclusion

Individual attitudes toward social insurance and redistribution depend on income, perceived unemployment risk, the possibilities of finding new employment—and the ability to self-insure. While many people can draw on savings to sustain levels of consumption in the event of negative economic shocks, large shares of the population, including many middle-class and rich individuals, have small or nonexistent economic buffers. This, we argue, has implications for the political economy of social insurance and redistribution.

Our main contribution is to show theoretically and empirically that being liquidity constrained, defined as lacking liquid savings and/or being unable to borrow in the credit market in order to maintain one’s standard of living, is associated with higher support for social insurance and income maintenance. While high-income citizens typically have larger savings in absolute terms, social insurance is about maintaining one’s standard of living, often dictated in the short term by committed consumption. This means that the relevant metric for assessing the adequacy of savings is the availability of funds relative to disposable income. At the same time, not all types of wealth are equally suited for alleviating temporary economic hardship. The middle and upper classes may, by definition, be wealthy, but often this wealth is illiquid and cannot, or only at great cost, be used to smooth consumption across economic shocks. For everyone except the very rich, illiquid wealth does not substitute for a lack of liquid savings, and liquidity constraints continue to be a predictor of support for UI across standard divisions of political conflict often cast in terms of rich versus poor.

Throughout, our data allow us to control for other recent explanations for individual support for social insurance and redistribution, including skill specificity, unemployment risk, and homeownership. We find support for all of these, and see the literature moving toward a broad understanding that individual economic circumstances in credit and labor markets are important determinants of social policy preferences. We concur with Hacker et al. (2013) that “too often the discussion of the economic roots of policy attitudes has declared—based on, at best, limited measures of dynamic economic experiences—that economic factors are largely irrelevant.” Indeed, the findings presented here underscore the need for improved measures of individual economic circumstances—as illustrated by the important distinction between illiquid and liquid wealth.

Our analyses are based on very detailed and accurate data from Denmark. These data allow for empirical investigations that would be difficult to carry out in most other countries. Our results can serve as a benchmark for analyses of the political implications of a fragile middle class in other countries and contexts, but the one-country nature of our primary data does not allow for analyses of the consequences of differences in credit markets or in traditions of informal insurance arrangements. In our case, a straightforward cross-country analysis based on proximate survey measures from the European Social Survey confirmed our main finding across a broad group of countries. However, to the best of our knowledge existing cross-country data sets rarely, if ever, combine detailed information on political and attitudinal variables with data on household balance sheets in the form of assets, debt, and interest rates. While the former are standard instruments in value and social surveys, the latter are typically confined to focused economic surveys. Combining insights from Iversen and Soskice (2001), Rehm, (2009, 2016), Hacker et al. (2013), Ansell (2014), and the present article would suggest the need for a novel cross-country data collection effort that integrates such information. This should be a priority to improve our understanding of the links between credit markets and financial regulation, patterns of labor market risk, informal safety nets, formal social insurance systems, and political institutions across industrialized countries.

Supplemental Material

data_and_programs – Supplemental material for Middle Class Without a Net: Savings, Financial Fragility, and Preferences Over Social Insurance

Supplemental material, data_and_programs for Middle Class Without a Net: Savings, Financial Fragility, and Preferences Over Social Insurance by Jacob Gerner Hariri, Amalie Sofie Jensen and David Dreyer Lassen in Comparative Political Studies

This article is distributed under the terms of the Creative Commons Attribution 4.0 License (http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/open-access-at-sage).

Supplemental Material

Liquidity_Online_Appendix_final_ddl – Supplemental material for Middle Class Without a Net: Savings, Financial Fragility, and Preferences Over Social Insurance

Supplemental material, Liquidity_Online_Appendix_final_ddl for Middle Class Without a Net: Savings, Financial Fragility, and Preferences Over Social Insurance by Jacob Gerner Hariri, Amalie Sofie Jensen and David Dreyer Lassen in Comparative Political Studies

Footnotes

Acknowledgements

We thank Jim Alt, Torben Iversen, Søren Leth-Petersen, Andreas Wiedemann, referees, and conference and seminar participants at EPSA, APSA, Harvard, Bergen, University College London, and the DPCW for comments. We gratefully acknowledge financial support from the Sapere Aude program under the Danish Independent Research Fund for the project “Hard Times, Hard Choices” and from an ERC Starting Grant (Project No. 313673). The activities of CEBI are financed through a grant from the Danish National Research Foundation.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: We gratefully acknowledge financial support from the Sapere Aude program under the Danish Independent Research Fund for the project “Hard Times, Hard Choices” and from an ERC Starting Grant (Project No. 313673). The activities of CEBI are financed through a grant from the Danish National Research Foundation.

ORCID iD

David Dreyer Lassen

Supplemental Material

Supplemental material for this article is available online at the CPS website

Notes

Author Biographies

Jacob Gerner Hariri is a professor (mso) at the Department of Political Science, University of Copenhagen. Jacob holds a PhD in political Science from the University of Copenhagen. His interests are in comparative politics with a focus on political regime developments, state formation, and social policy preferences.

Amalie Sofie Jensen is a postdoctoral research associate in the Princeton Program on Public Finance and the Woodrow Wilson School of Public and International Affairs. Amalie holds a PhD in economics from the University of Copenhagen. Her research interests are in public economics and political economy with a focus on taxation and transfers, household consumption and savings, and voter preferences for economic policies.

David Dreyer Lassen is a professor of Economics at the University of Copenhagen. He holds a PhD in economics, is the co-deputy director of the Center for Economic Behavior and Inequality, and the director of the Center for Social Data Science. His research interests focus on political economy and social data science.

References

Alesina

Angeletos

(2005). Fairness and redistribution. The American Economic Review, 95, 960-980.

Alt

J. E.

Barfort

Lassen

D. D.

(2017, October). The effects of income and unemployment shocks on political preferences. Paper presented at the 2017 NBER Political Economy Conference. Cambridge, MA, USA.

Ansell

(2014). The political economy of ownership: Housing markets and the welfare state. American Political Science Review, 108, 383-402.

Atkinson

A. B.

Brandolini

(2011). On the identification of the “middle class.” In Gornick

J. C.

Jantti

(Eds.), Income inequality: Economic disparities and the middle class in affluent countries (pp. 77-100). Stanford, CA: Stanford University Press.

Baldwin

(1990). The politics of social solidarity: Class bases of the European welfare state, 1875-1975. Cambridge, MA: Cambridge University Press.

Basten

Fagereng

Telle

(2014). Cash-on-hand and the duration of job search: Quasi-experimental evidence from Norway. The Economic Journal, 124, 540-568.

Basten

Fagereng

Telle

(2016). Saving and portfolio allocation before and after job loss. Journal of Money, Credit and Banking, 48, 293-324.

Card

Chetty

Weber

(2007). Cash-on-hand and competing models of intertemporal behavior: New evidence from the labor market. The Quarterly Journal of Economics, 122, 1511-1560.

Chetty

Szeidl

(2007). Consumption commitments and risk preferences. The Quarterly Journal of Economics, 122, 831-877.

10.

Chiuri

M. C.

Jappelli

(2010). Do the elderly reduce housing equity? An international comparison. Journal of Population Economics, 23, 643-663.

11.

Cusack

Iversen

Rehm

(2006). Risks at work: The demand and supply sides of government redistribution. Oxford Review of Economic Policy, 2, 365-389.

12.

Davis

Zhu

(2011). Bank lending and commercial property cycles: Some cross-country evidence. Journal of International Money and Finance, 30, 1-21.

13.

Duch

R. M.

Rueda

(2016). The people you are: Personality traits as determinants of redistribution preferences (Working Paper). Retrieved from https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2840433

14.

Engen

E. M.

Gruber

(2001). Unemployment insurance and precautionary saving. Journal of Monetary Economics, 47, 545-579.

15.

Esping-Andersen

(1990). The three worlds of welfare capitalism. Princeton, NJ: Princeton University Press.

16.

Esping-Andersen

Korpi

(1984). Social policy as class politics in post-war capitalism: Scandinavia, Austria, and Germany. In Goldthorpe

(Ed.), Order and conflict in contemporary capitalism (pp. 179-208). Oxford, UK: Clarendon Press.

17.

Estevez-Abe

Iversen

Soskice

(2001). Social protection and the formation of skills: A reinterpretation of the welfare state. In Hall

P. A.

Soskice

(Eds.), Varieties of capitalism: The institutional foundations of comparative advantage (pp. 145-183). Oxford, UK: Oxford University Press.

18.

Friedman

(1957). A theory of the consumption function. Princeton, NJ: Princeton University Press.

19.

Gingrich

Ansell

(2012). Preferences in context: Micro preferences, macro contexts, and the demand for social policy. Comparative Political Studies, 45, 1624-1654.

20.

Gruber

(2001). The wealth of the unemployed. Industrial and Labor Relations Review, 55, 79-94.

21.

Hacker

J. S.

Rehm

Schlesinger

(2013). The insecure American: Economic experiences, financial worries, and policy attitudes. Perspectives on Politics, 11, 23-49.

22.

Hariri

J. G.

Lassen

D. D.

(2017). Income and outcomes: Social desirability bias distorts measurements of the relationship between income and political behavior. Public Opinion Quarterly, 81, 564-576.

23.

Hendren

(2017). Knowledge of future job loss and implications for unemployment insurance. The American Economic Review, 107, 1778-1823.

24.

Iversen

Soskice

(2001). An asset theory of social preferences. American Political Science Review, 95, 875-893.

25.

Johnson

D. S.

Parker

J. A.

Souleles

N. S.

(2006). Household expenditure and the income tax rebates of 2001. The American Economic Review, 96, 1589-1610.

26.

Kaplan

Violante

G. L.

Weidner

(2014). The wealthy hand-to-mouth (Brookings papers on economic activity). Cambridge, MA: National Bureau of Economic Research.

27.

Kautto

Fritzell

Hvinden

Kvist

Uusitalo

(2001). Nordic welfare states in the European context. London, England: Routledge.

28.

Korpi

(1983). The democratic class struggle. London, England: Routledge & Kegan Paul.

29.

Kreiner

C. T.

Lassen

D. D.

Leth-Petersen

(2019). Liquidity constraint tightness and consumer responses to fiscal stimulus policy. American Economic Journal: Economic Policy, 11, 351-379.

30.

Leth-Petersen

(2010). Intertemporal consumption and credit constraints: Does total expenditure respond to an exogenous shock to credit? The American Economic Review, 100, 1080-1103.

31.

Lusardi

Schneider

Tufano

(2011). Financially fragile households: Evidence and implications. Brookings Papers on Economic Activity, 1, 38-134.

32.

Meltzer

A. H.

Richard

S. F.

(1981). A rational theory of the size of government. Journal of Political Economy, 89, 914-927.

33.

Moene

K. O.

Wallerstein

(2001). Inequality, social insurance, and redistribution. American Political Science Review, 95, 859-874.

34.

Moene

K. O.

Wallerstein

(2003). Earnings inequality and welfare spending: A disaggregated analysis. World Politics, 55, 485-516.

35.

Rehm

(2009). Risks and redistribution: An individual-level analysis. Comparative Political Studies, 42, 855-881.

36.

Rehm

(2011). Risk inequality and the polarized American electorate. British Journal of Political Science, 41, 363-387.

37.

Rehm

(2016). Risk inequality and welfare states: Social policy preferences, development, and dynamics. Cambridge, MA: Cambridge University Press.

38.

Sullivan

T. A.

Warren

Westbrook

J. L.

(2000). Fragile middle-class: Americans in debt. New Haven, CT: Yale University Press.

39.

The Unemployment Benefits Committee. (2015). Privat forsikring, supplerende ydelser og fratrædelsesgodtgørelser [Private insurance, supplementary benefits, and severance pay]. Copenhagen, Denmark: Danish Ministry of Employment.

40.

Warren

Tyagi

A. W.

(2003). The two-income trap. New York, NY: Basic Books.

41.

Zeldes

S. P.

(1989). Consumption and liquidity constraints: An empirical investigation. Journal of Political Economy, 97, 914-992.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

16.31 MB

0.00 MB

0.70 MB